Multimodal LLM, Geospatial AI, and Urban Policy: Estimating Redlining's Legacy effects using Street View Imagery

These replication data are for the paper: MLLMs, Street View and Urban Policy-Intelligence: Recovering the Sustainability Effects of Redlining. 

The paper evaluates whether multimodal large language models (MLLMs) can derive neighborhood-level sustainability indicators from Google Street View (GSV) imagery and recover the legacy effects of historical redlining in the Phoenix metropolitan area. We compare MLLM-based inference (GPT-4o) against conventional semantic segmentation (ResNet-based) and authoritative benchmarks (ACS poverty rates, GEIE tree canopy coverage).